STRATEGY. TACTICS. TRADECRAFT.
I build systems for testing, containing and hardening frontier AI agents. I lead adversarial operations and advise teams deploying them.
Explore my work01 / CURRENT WORK
An agent acts on what it believes about the world. I study how its behavior changes when we vary what it sees, who it trusts and whether it believes it has succeeded. Its actual permissions remain independently controlled.
FOUNDER & CEO
Adaptive environments for frontier AI.
I’m building HyperRange: generated environments, controlled differences between actor views and scenarios that evolve during execution.
It extends a lab’s evaluation harness to investigate coordinated agents, test containment and follow behavior beyond apparent success.
Explore HypergameMANAGING DIRECTOR
Frontier AI red team operations.
I lead sustained adversarial campaigns against models, agents and their infrastructure, working with internal security and research teams.
I shape campaigns, assemble operators and stay close to the findings that drive engineering changes.
Explore BT602 / SPEAKING
Technical talks, live demonstrations and private briefings on AI security, autonomous systems and adversarial operations.
BLACK HAT USA 2026
Agent Compromise With a Physical Blast Radius
Embodied AI security, prompt injection and physical consequences. A BT6 briefing with a live robot demonstration.
View the briefingSTANFORD / SCALING TRUST MEETUP
Live demonstration and discussion at the Scaling Trust Meetup.
THE FIELD MANUAL / NO STARCH PRESS
A Field Manual for Attacking
Intelligent Systems
How to reason about an AI attack surface, design an adversarial campaign and turn technical findings into a useful security decision.
Get the bookEarly-access ebook and print preorder available.
Print edition listed for Spring 2027.
03 / WRITING
I founded AI Security Pro to publish technical work on adversarial systems, cyber deception and adaptive defense.
Visit AI Security ProHypergame theory, misperception and autonomous threat engagement.
Deception, changing environments and the case for adaptive defense.
The dependencies beneath an AI system, from data and software to infrastructure.
04 / ADVISORY
Independent technical advice for founders, security leaders and teams deploying AI into consequential workflows.
Define the threat model, identify attack paths and turn findings into engineering priorities.
Threat models · Test strategy · Mitigation review
Establish what an agent can reach, which controls enforce the boundary and how to test whether they hold.
Architecture review · Execution boundaries · Evidence
Assess products, architectures and technical claims. Determine what the evidence supports and what needs investigation.
Founder advisory · Technical diligence · Leadership briefings
Advisory and publishing are offered through AI Security LLC.
Hands-on red team operations and HyperRange studies are scoped through BT6 and Hypergame respectively.
05 / BACKGROUND
I’ve spent nearly two decades engineering secure systems, building companies and leading adversarial operations. Across three AI ventures and my work as a CISO, the central question has been how to build systems that hold up against an adversary who learns and adapts.
My research with Dr. Mark Mateski on hypergame theory and deception began in 2016. It gave that work a sharper focus: how an adversary’s understanding of the environment shapes its decisions, and how defenders can influence that understanding.
I brought that perspective to AI infrastructure as co-founder and CISO of Hydra Host. Today, I’m advancing it through Hypergame’s adaptive environments and BT6’s frontier AI red team operations: testing what agents perceive, what they attempt and whether the controls around them hold.
GET IN TOUCH
Tell me what you’re building and the problem you want to solve.
phil@aisecurity.proResearch partnerships, advisory, speaking and publishing.
Start with a non-sensitive summary.